Blood gene expression, lifestyle and diet The Norwegian Women and Cancer Post-genome Cohort

(1)

FACULTY OF HEALTH SCIENCES

DEPARTMENT OF COMMUNITY MEDICINE

Blood gene expression, lifestyle and diet

The Norwegian Women and Cancer Post-genome Cohort

Karina Standahl Olsen

A dissertation for the degree of Philosophiae Doctor

2013

(2)

(3)

Blood gene expression, lifestyle and diet

The Norwegian Women and Cancer Post-genome Cohort

Karina Standahl Olsen

Department of Community Medicine Faculty of Health Sciences

University of Tromsø Tromsø, Norway

2013

(4)

(5)

Acknowledgements

The work presented here was conducted at the Department of Community Medicine (ISM), Faculty of Health Sciences, University of Tromsø, from the autumn of 2006 to the spring of 2013. The work was sponsored by a university grant. Many people contributed more or less directly and at different stages to the completion of this thesis. I am grateful to all of you.

My main supervisor Eiliv Lund: Thank you for sharing your ideas and visions, and for making a (back then) young PhD student feel that her opinion and knowledge matters. Your guidance during these years, and your patience and understanding during hard times has been

invaluable.

My co-supervisor Ruth H. Paulssen: Thank you for taking the time to talk about molecular biology, life, kids, and cats. I am also grateful for your comments to the thesis drafts.

Co-authors: your contributions and insightful comments have been highly valued, and your knowledge in your respective fields of research have taught me a lot.

Administrative staff at NOWAC and ISM, especially Bente, Merete, and Mona: thank you for fixing everything that needs to be fixed. Your magic makes life at ISM go around!

My office mates over the years: First, Kristin S.: I am so glad to have had you as my skillful and knowledgeable partner in crime when it comes to all things lab-related. Our long talks on all aspects of lab & life have meant a lot to me, you are a gem. Also, Marit, Oxana, Lotta and Sara:

thank you for insights into life and science, and for listening to spontaneous outbursts of joy and sorrow as the ups and downs of life as a PhD student have unfolded.

Other colleagues, especially Kristin B., Guri, Toril, Tor Gisle, Bent-Martin, Kristin (snipis!), and many more: thank you for making walking up and down the hallways at ISM all day,

(6)

every day, a very nice activity, and for being the friendly faces one needs to see every now and then. The work-out group: thank you for the inspiration to postpone any potential

musculoskeletal consequences of long hours by the pc.

It should be noted that I never had the intention of working with fatty acids: their biology is just too much to handle. However, my dear friend Caroline encouraged me to join in the research group of Svanhild Schønberg, to do my MSc together with her. Svanhild introduced me to the PUFAs, and she has been an inspiration. So has Caroline, who sparked my

enthusiasm for microarrays. I am grateful to you both.

Friends outside work, especially my girls in Trondheim and Tromsdalen, as well as Reidun, Lene, Yngvild, and Ingeborg: you are all unique in your own special ways and you add great value to my life. Also, to my yoga teacher Torgunn: your classes are a weekly source of inspiration.

My family and family-in-law: without your undivided love and support my life would have been very different, and this project would never have come to a completion. In particular, thank you, mamma, for always encouraging me and being there for me.

Finally, to my dear husband Kenneth: I am deeply grateful for your patience and efforts during my years as a PhD student, and especially during the completion of my thesis. Du e min klippe. Mia: mamma sitt gull, thank you for putting everything into perspective, just by being here.

Karina

Tromsø, June 2013

(7)

Summary

Lifestyle and diet are risk factors for cancer, cardiovascular and inflammatory diseases, but the molecular mechanisms involved have not been fully elucidated. High-throughput

technologies like microarrays are being used to explore those mechanisms, but data from model systems (cell cultures and research animals) may not reflect human biology as well as previously thought.

The aim of this PhD thesis was to explore the feasibility of using blood samples from a population-based cohort to generate gene expression profiles that enable identification of molecular mechanisms related to several lifestyle and dietary factors.

The Norwegian Women and Cancer (NOWAC) study is a nationally representative, prospective cohort started in 1991 to investigate risk factors for breast cancer. The study includes more than 172 000 women who have answered questionnaires regarding lifestyle, diet, and health. In 2003-06, a subset of women in NOWAC were randomly drawn to constitute the NOWAC Post-genome Cohort, and approx. 50 000 blood samples eligible for gene expression analysis were collected, along with an additional blood sample and a

questionnaire. For this thesis, full-genome gene expression microarrays, as well as plasma concentrations of fatty acids and vitamin D were analyzed in 500 samples from the Post- genome Cohort. For paper I, gene expression related to major lifestyle-related exposures (smoking, medication use), inter-individual variation (age, body mass index, fasting status), and technical factors (pre-analytical and analytical) was explored. Paper II and III took a nutrigenomics approach to explore gene expression related to vitamin D and fatty acid ratios.

Due to the partially overlapping dietary source (fatty fish), the co-variability of vitamin D and fatty acids were analyzed in paper IV.

Technical variability influenced the gene expression profiles considerably, pointing to the need for rigorous pre-analytical and analytical control. In addition, inter-individual and exposure variables were associated with differential gene expression, both at the gene and pathway level. Vitamin D status was associated with modest gene expression differences.

However, at the pathway level, several immuneregulatory processes were identified, supporting the emerging hypotheses of vitamin D as an immuneregulatory nutrient. The ratios of different polyunsaturated fatty acids of the n-6 and n-3 families were associated with both gene- and pathway-level differences, which pointed to cellular and molecular

mechanisms influenced by dietary fat. The concentrations of vitamin D and the investigated fatty acid ratios were only weakly associated, indicating that the identified gene expression profiles likely arise from the unique influence of the two factors separately.

In conclusion, the thesis demonstrates that several exposures related to lifestyle and diet, as well as technical variability, are mirrored in blood gene expression profiles. This implies that gene expression analysis using blood samples may become an important tool for identification of molecular mechanisms related to many exposures, for characterization of pathological processes related to cancer and other diseases, and for prediction of disease risk.

(10)

Sammendrag

Livsstil og kosthold er risikofaktorer for kreft, kardiovaskulære og inflammatoriske

sykdommer, men hvilke molekylære mekanismer som er involvert er ikke fullt ut beskrevet.

Teknologi som microarray (mikromatriser) brukes i stor grad for å utforske disse

mekanismene, men resultater fra modellsystemer som cellekulturer og forsøksdyr reflekterer ikke nødvendigvis menneskets biologi så godt som man tidligere har trodd.

Målet med denne doktorgradsavhandlingen var å utforske muligheten for å bruke blodprøver fra en populasjsonsbasert kohorte, for å analysere genuttrykksprofiler og dermed identifisere molekylære mekanismer relatert til ulike livsstils- og kostholdsfaktorer.

Kvinner og Kreft-studien er en nasjonalt representativ, prospektiv kohorte som ble startet i 1991, for å forske på risikofaktorer for brystkreft. Studien inkluderer i dag mer enn 172 000 kvinner, som har svart på spørreskjema angående livsstil, kosthold, og helse. I 2003-06 ble ca.

50 000 av deltakerne tilfeldig plukket ut for å delta i Kvinner og Kreft Post-genom kohorten.

Fra disse kvinnene ble det samlet inn blodprøver med mulighet for genuttrykksanalyse, samt en standard blodprøve og et spørreskjema. Til denne avhandlingen ble det brukt materiale fra 500 kvinner i Post-genom kohorten. Fullgenom genuttryksanalyser ble uført, i tillegg til måling av konsentrasjon av fettsyrer og vitamin D i plasma. I artikkel I ble genuttrykk relatert til livsstilseksponeringer (røyking, medisinbruk), inter-individuell variasjon (alder,

kropsmasseindeks, fastestatus), og tekniske faktorer (pre-analytiske og analytiske) utforsket. I artikkel II og III ble genuttrykk relatert til vitamin D og fettsyreratioer analysert, og disse artiklene kan sies å være en del av forskningsfeltet “nutrigenomics”. På grunn av deres delvis overlappende kostholdskilde, fet fisk, ble samvariasjon mellom vitamin D og fettsyreratioer analysert i artikkel IV.

Teknisk variasjon påvirket genuttrykksprofilene i stor grad, noe som peker på behovet for tett kontroll av preanalytiske og analytiske prosedyrer. Inter-individuelle faktorer og ulike

eksponeringer var assosiert med differensielt uttrykk både av enkeltgener og av signalspor.

Vitamin D-status var assosiert med beskjedne forskjeller i genuttrykk. Likevel fant vi, på signalspornivå, flere immunologiske prosesser. Dette støtter hypotesen om at vitamin D kan være immunregulatorisk. Ratioen mellom flerumettede fettsyrer tilhørende n-6 eller n-3- familiene var assosiert med differensielt uttrykk av både enkeltgener og signalspor, og pekte ut prosesser som kan bli påvirket av ulike typer fett fra kosten. Konsentrasjonene av vitamin D og fettsyreratioer var kun svakt assosiert med hverandre, noe som indikerer at

genuttrykksprofilene oppstår som følge av uavhengig påvirkning fra disse to kostholdsfaktorene.

Funnene i avhandlingen viser at ulike typer eksponeringer relatert til livsstil og helse, i tillegg til teknisk variasjon, gjenspeiles i genuttrykksprofiler i blod. Dette antyder at

genuttrykksanalyser i blodprøver kan bli et viktig verktøy for å identifisere molekylære

mekanismer relatert til ulike typer eksponeringer, for å karakterisere sykdomsprosesser, og for å forutsi risiko for sykdom.

(11)

List of papers

This thesis is based on the following papers, hereafter referred to by their roman numerals.

Paper I

Dumeaux V, Olsen KS, Nuel G, Paulssen RH, Borresen-Dale AL, Lund E (2010). Deciphering normal blood gene expression variation--The NOWAC postgenome study. PLoS Genet 6(3): e1000873. PubMed ID: 20300640

Paper II

Olsen KS, Rylander C, Brustad M, Aksnes L, Lund E (2013). Plasma 25 hydroxyvitamin D level and blood gene expression profiles: a cross-sectional study of the Norwegian Women and Cancer Post-genome Cohort. Eur J Clin Nutr. Advance online publication 6 March 2013.

PubMed ID: 23462941

Paper III

Olsen KS, Fenton C, Frøyland L, Waaseth M, Paulssen RH, Lund E (2013). Plasma fatty acid ratios affect blood gene expression profiles – A cross-sectional study of the Norwegian Women and Cancer Post-genome Cohort. PLOS ONE [accepted]

Paper IV

Olsen KS, Aksnes L, Frøyland L, Lund E, Rylander C. Vitamin D status and PUFA ratios in a national representative cross-section of healthy middle-aged, Norwegian women

[manuscript]

(12)

List of figures and tables

Figure 1: Overview and examples of biomarkers 10

Figure 2: The structure of alpha-linolenic acid, 18:3n-3 13

Figure 3: Overview of PUFA metabolism and eicosanoid production 15 Figure 4: Major pathways of fatty acid production, transport, and metabolism 21

Figure 5: Overview of vitamin D metabolism 23

Figure 6: The structure of calcitriol (1,25(OH)2D) 23

Figure 7: Central dogma of molecular biology 29

Figure 8: Basic principle of hybridization-based microarrays 30 Figure 9: Mechanisms by which altered supply of fatty acids could affect immune responses 62 Figure 10: The globolomic design of the NOWAC Post-genome Cohort, compared to a

traditional prospective study design. 70

Table 1: Reports on aspects of nutrition and lifestyle 11

Table 2: Recommended and achieved intake of fat and PUFAs 16

(13)

Abbreviations

AA Arachidonic acid

ABCG1 ATP-binding cassette sub-family G member 1 ABI Applied Biosystems

AICR American Institute of Cancer Research ALA Alpha-linolenic acid BMI Body mass index

COX Cyclooxygenase

CV Coefficient of variation CVD Cardio-vascular diseases DBP Vitamin D binding protein

DHA Docosahexaenoic acid DNA Deoxyribonucleic acid EPA Eicosapentaenoic acid

FA Fatty acid

FABP4 Fatty acid binding protein 4

FAO Food and Agriculture Organization of the United Nations GO Gene Ontology

GSEA Gene Set Enrichment Analysis

HPLC High-performance liquid chromatography IL Interleukin

IOM Institute of Medicine

KEGG Kyoto Encyclopedia of Genes and Genomes LA Linoleic acid

LOX Lipoxygenase

MAQC Microarray Quality Control project mRNA Messenger ribonucleic acid

MSigDB Molecular Signatures Database NF-κB Nuclear factor κB

NOWAC Norwegian Women And Cancer study PI3K Phosphatidylinositol-3-kinase PLC Phospholipase C

PPAR Peroxisome proliferator-activating receptor PUFA Polyunsaturated fatty acid

qRT-PCR Quantitative real-time reverse transcription polymerase chain reaction RNA Ribonucleic acid

RXR Retinoid x-receptor TLR Toll-like receptor TNF Tumor necrosis factor UVB Ultraviolet B VDR Vitamin D receptor WCRF World Cancer Research Fund WHO World Health Organization

(14)

1 Introduction

The human genome was characterized around the year 2000, marking the start of the era of genomic research. After the sequencing of the genome, large initiatives were launched to identify potentially disease-causing genomic variation. So far, results from those initiatives have been modest. An alternative focus for genomics research has been the study of gene activity, known as gene expression. With gene expression analysis using full-genome microarrays, the molecular mechanisms involved in the association between lifestyle, diet, health and disease can be approached with a genome-wide perspective. However, gene expression analysis has not been embraced in large-scale, population-based studies to the same extent as the study of gene variation.

To explore this research gap, an extensive collection of blood samples from a representative selection of the Norwegian female population was initiated in the Norwegian Women and Cancer (NOWAC) Post-genome Cohort. Unlike the samples collected by most other cohort studies, these blood samples were eligible for gene expression analysis. The present thesis takes advantage of this unique sample material, to explore potential associations between gene expression profiles and several exposures including lifestyle and dietary factors, in a cross- sectional study design. Due to the fact that Norwegians consume large amounts of fish compared to many other countries, two nutrients related to fish (polyunsaturated fatty acids (PUFAs) and vitamin D) are the focus of three of the papers included.

1.1 Lifestyle and nutrition in the etiology of human disease

1.1.1 Historical aspects and potential for disease prevention

Throughout human history, lifestyle and dietary patterns have been changing [1]. The

transition from hunter-gatherer societies, via the peasant-agricultural, to the urban-industrial societies has led to a diet that is increasingly based on animal products, and that contains more energy, more fats from plants and animals, and more sugar. The consumption of fish today is highly variable according to geography [2] and economy [3]: rich countries consume

(15)

more fish than poorer countries. Also, different kinds of fish and seafood are consumed, and they provide varying amounts of nutrients [2]. The most recent steps of the dietary transition have coincided with a drop in human energy requirements, as the change of lifestyle

occurring in industrialized societies during the last 60 years has led to a reduction in overall physical activity [1].

Dietary patterns and other lifestyle factors related to historical, geographical, and (socio-) economic changes, have led to increased life expectancy, but these changes also correlate with global disease patterns [1]. For example, during the last 50 years, European consumption of fat has increased by approx. 50%, and prevalence of obesity in adults has doubled [4]. Also, the incidence of lifestyle-related cancers and cardiovascular disease (CVD) increases in people that adopt a so-called Western lifestyle [1]. These examples point to the contribution of modifiable factors to disease etiology, and also highlight the potential for disease prevention.

It has been estimated that approximately one quarter of cancer cases in high income countries can be prevented through adopting a healthy eating pattern, maintaining a healthy weight, and being physically active [1]. Based on an extensive review of current evidence, the World Cancer Research Fund (WCRF) and the American Institute for Cancer Research (AICR) formulated eight main recommendations for healthy living, with the aim of reducing cancer risk [1]. It has now been shown that following these recommendations may lower overall cancer risk by 18%, with numbers differing according to cancer site [5]. For example, breast cancer risk may be reduced by 16%, colorectal cancer by 27%. Furthermore, adherence to the WCRF/AICR recommendations was associated with significantly increased longevity, as well as reduced hazard of death caused by cancer, cardiovascular diseases (CVD), and respiratory diseases [6]. In addition to the factors related to diet, body fatness and physical activity, smoking is the major lifestyle-related risk factor for cancer. Tobacco use is estimated to cause 22% of cancer deaths worldwide, making it the greatest single avoidable risk factor [7].

Moreover, smoking is an even more important risk factor for CVD and respiratory diseases [8].

(16)

1.1.2 Tools for research on nutrition and lifestyle

Although the associations and biological mechanisms for the influence of tobacco smoke on disease has been extensively evaluated and reported, this is not the case with some of the other risk factors mentioned above, including diet, weight, and physical activity. Epidemiology may find associations between exposures and diseases, but there is a demand for biological

plausibility for the associations to be regarded as potentially causal. Many dietary constituents are being evaluated as potential risk (or protective) factors for disease, but as a typical diet may provide more than 25 000 bioactive constituents, the characterization of associations and potential mechanisms is a complex task [1]. Important tools in the research for causal

mechanisms related to nutrition and lifestyle factors include the use of biomarkers, and increasingly, the use of high-throughput molecular biology technologies. Biomarkers may be defined as any substance, structure or process that can be measured in the body, or its

products, that influences or predicts the incidence of disease [9]. Generally, they are classified into biomarkers of exposures and biomarkers of effects (Figure 1). High-throughput

technologies have evolved as an effort to elucidate the interaction of exposures and cellular and physiological homeostasis [1]. Gene expression profiles may be regarded as biomarkers of early effects (Figure 1) [10], and the technology will be further introduced in chapter 1.4.

Figure 1: Overview and examples of biomarkers. Adapted from [9].

A main goal of medical research and epidemiology is disease prevention. Despite decades of research, more knowledge is needed at both the epidemiological and molecular level to describe and thereby be able to identify pathogenic processes in a pre-clinical setting. Only

(17)

then can interventions be made at the individual level, and preventive measures can be taken in the form of lifestyle and dietary advice to the public. The public health perspective is reflected by the fact that governments and institutions initiate the publication of large reports that describe the state of the available evidence, and attempt to draw conclusions based on reported findings. Several such reports are referred to in this thesis (Table 1).

Table 1: Reports on aspects of nutrition and lifestyle.

Title of report Institution Short name Ref.

Food, nutrition, physical activity, and the prevention of cancer: a global perspective

World Cancer Research Fund (WCRF) and American Institute of Cancer Research (AICR)

WCRF/AICR report

[1]

Fats and fatty acids in human nutrition: report of an expert consultation

World Health Organization (WHO) and the Food and Agriculture Organization of the United Nations (FAO)

WHO/FAO report

[11]

A comprehensive assessment of fish and other seafood in the Norwegian diet

Norwegian Scientific Committee for Food Safety

- [12]

Dietary Reference Intakes for Calcium and Vitamin D

Institute of Medicine (IOM), funded by National Institute of Health

IOM report [13]

It should be mentioned that research on effects of nutrients may complicated by non-linear dose-response relationships. Nutrients may display a U-shaped risk curve, where both too low and too high levels are associated with adverse effects [14]. This complicates interpretation of reported findings. This property must also be taken into account when issuing dietary advice to the public: excessive intake, for example through dietary supplements, may be harmful.

However, excessive intake will not be further discussed here, as the general tendency in the Norwegian population is deficiency of the nutrients in question.

1.1.3 Fish consumption

Fish is considered part of a healthy diet and provides a number of macro- and micronutrients that may be important for disease etiology. These nutrients include protein, PUFAs, retinol, vitamins D and E, selenium and iodine. Of these, PUFAs and vitamin D are the focus of the

(18)

present thesis. On average, Norwegians consume approx. 55-65 grams of fish per day,

equivalent to approx. two fish meals per weak [12, 15]. This provides 40% of ingested vitamin D per day, and 9% of ingested PUFAs per day [15]. In a study of ten European countries, Norwegians consumed the second highest amounts of fish [2]. However, only 21% of women eat the recommended 200 grams of fatty fish per week (salmon, trout, mackerel, tuna,

sardines, herring) [15]. Compared to other nationalities, Norwegians eat high amounts of very fatty fish, and high amounts of white fish products [2]. Consumption differs according to population strata, with potential adverse consequences for nutrient status in low-intake population strata [12]. Increased fish consumption is encouraged by Norwegian health authorities, but no specific amount is recommended [16].

The protective effect of fish consumption for CVD, inflammatory diseases, and possibly cancer is primarily linked to n-3 PUFAs which are present mainly in fatty fish. Also, vitamin D is present in higher amounts in fatty fish. Hence, it may not be reasonable to expect clear associations between total fish consumption and diseases. In the WCRF/AICR report, the available evidence for a potential association of cancer with consumption of fish was evaluated [1]. The report concluded that there is “limited-suggestive” evidence for a reduction in risk of colorectal cancer. An important limitation of the included studies was the broad definition of fish, with no stratification into lean and oily fish. Similarly, a recent systematic review, which included data from 106 case-control and cohort studies, concluded that the hypothesis that fish consumption reduces risk of prostate, breast and colorectal cancers is supported by little epidemiological data [17]. However, several challenges were pointed out by the authors, including the fact that many studies had not stratified between lean and fatty fish. Studies that did stratify, found inverse relationships between fatty fish and the risk of breast [18] and prostate cancer [19]. However, these association have not been found by all [20]. Fish is stored and processed in a number of different ways that may alter the properties of PUFAs, or

introduce harmful substances such as high amounts of salt, or mutagenic compounds from frying. Storage and processing may modify the potential association of fish consumption and cancer, and consumption of pure filet was found to provide the most favorable PUFA

(19)

composition [21]. Other potential confounding factors and limited sample sizes must be taken into consideration when evaluating the literature.

1.2 Fatty acids

1.2.1 Structure and metabolism

Fatty acids (FAs) are macronutrients that are involved in multiple cellular and physiological mechanisms. FAs consist of a hydrocarbon chain with a methyl group (CH3) at one end (denoted as the omega- or n-end) and a carboxyl group (COOH) at the other end (Figure 2).

Chain length typically varies from 14 to 24 carbon atoms, and the presence and number of double bonds gives rise to the classification into saturated (SFA, no double bonds and

therefore saturated with hydrogen atoms), monounsaturated (MUFA, one double bond), and PUFAs (multiple double bonds). FAs are named according to chain length, number of double bonds, and position of the first double bound counted from the n-end (Figure 2). The

presence of double bonds alters the physical properties of the FA, by introducing kinks in the carbon chain, and rendering the FAs more prone to oxidation.

Figure 2: The structure of alpha-linolenic acid, 18:3n-3 (PubChem CID 5280934). Reprinted with permission [22].

FAs from the diet or from lipolysis in adipose tissue, enter the blood stream esterified in lipoproteins such as chylomicrons (dietary FAs) or very low density lipoproteins (VLDL,

(20)

endogenous FAs), or non-esterified bound to albumin. FAs are transported into cells throughout the body, where they may be stored as triglycerides, incorporated into the phospholipids of cellular membranes, modified by chain elongation and desaturation,

oxidation and peroxidation, and lastly, they serve as precursors for immunoactive eicosanoids (including prostaglandins, thromboxanes, leukotrienes, resolvins and others). In addition, FAs are potent regulators of gene expression, as described below [23, 24].

The major FAs provided through the diet include those of the n-3, n-6 and n-9 families.

Because mammals lack the enzyme that catalyses insertion of double bonds in the n-6 and n-3 position, two FAs are essential to humans and must be obtained through the diet: linoleic acid (LA, 18:2n-6) and alpha-linolenic acid (ALA, 18:3n-3). These serve as precursors for longer FAs with a higher degree of unsaturation, like the n-6 arachidonic acid (AA, 20:4n-6), and the n-3s eicosapentaenoic acid (EPA, 20:5n-3) and docosahexaenoic acid (DHA, 22:6n-3)

(Figure 3). In humans, this metabolism is limited [25]. Importantly, the classes of n-6 and n-3 FAs are not inter-convertible [26], but they compete for the same set of metabolic enzymes during chain elongation and desaturation. Due to the high content of LA in the Western diet, the n-6 FAs dominate the enzymatic conversions, resulting in increased amounts of AA. On the other hand, ALA has a higher affinity for the Δ6 desaturase, but is generally present in lower amounts. ALA is the precursor for EPA and DHA, but these are generally provided through the diet. The dietary intake of specific PUFAs will lead to the distribution of those PUFAs into virtually every cell in the body [27]. Intracellularly, the PUFAs are further metabolized to yield the eicosanoids, which may have differing or even opposite biological effects (Figure 3). In general, eicosanoids derived from AA are regarded as pro-inflammatory, as further discussed in chapter 1.2.4.

(21)

Figure 3: Overview of PUFA metabolism and eicosanoid production. Full names are stated in the list of abbreviations.

1.2.2 Sources and recommendations

Table 2 shows recommended and achieved intakes for total fat, various FA classes, and single FAs. In general, Norwegians consume higher amounts of fat than recommended, but lower amounts of PUFAs than recommended [15]. The consumption of marine PUFAs EPA and DHA was found to be just within the acceptable range, but the WHO/FAO report stated that the minimum value of 2 g/day for EPA/DHA may be increased in the future [11].

(22)

Table 2: Recommended and achieved intake of fat and PUFAs ^a

Recommended per day Achieved per day

% energy ^b Amount (g) % energy ^b Amount (g)

Total fat 20-35 (25-35)^N 30-34 67-74

PUFA ^c 6-11 (5-10)^N 5,3-6.2 13

Total n-6 2.5-9

Total n-3 ^d 0.5-2

LA ^e 2.5-9 8.8

ALA >0.5 1-2 1.2

EPA+DHA 0.25-2 0.35 0.84

a) Where values differ, values for women are given, and values for men are higher. Full names are stated in the list of abbreviations. Data from [11, 15, 28, 29].

b) % of total energy intake, minimum level or acceptable range.

c) LA+ALA+EPA+DHA.

d) ALA+ long chain n-3 FAs.

e) AA is not essential to adults whose diet provides >2.5 % energy of LA.

N) Recommendations in Norway.

N-6 PUFAs like LA are provided mainly from plant oils (corn, safflower, olive, peanut, and sunflower). The n-3 PUFA ALA is mainly provided in the form of plant oils from linseed, canola (modified rapeseed oil), and walnuts. Fatty fish and seafood are important sources of the long-chain n-3 PUFAs EPA and DHA. Smaller amounts of PUFAs are also present in meat, eggs and poultry, and processed foods that are common in the Western society are frequently rich in n-6 PUFAs. The most important sources of PUFAs in the Norwegian diet is butter, margarine, mayonnaise, dressings and bread [15]. Initiatives to increase n-3

consumption include dietary recommendations to increase consumption of (fatty) fish and oils from n-3 rich plants, such as linseed and rapeseed. In addition, Norwegian health authorities recommend the use of dietary supplements like n-3 oils/capsules or cod liver oil, and total sales for these supplements are 689 million NOK per year in Norway [30]. Among Norwegian women, 45 % take cod liver oil, but about half of the users take it only during the winter [31]. The standard dose of 5 ml cod liver oil (in the form of oil or capsules) provides 1.2 g n-3 PUFAs. However, a study found that only 58% of women taking cod liver oil took this amount [29]. Oils and capsules other than cod liver oil providing long chain PUFAs of varying kinds are also common in Norway. However, these have varying n-3 content, and do not contain vitamin D.

(23)

Recommendations have been put forward that the ratio of n-6/n-3 PUFAs should be

“balanced”, meaning 1:1, mainly based on the hypothesis that this was the dietary ratio during human evolution: our genes were evolutionary adapted to a balanced ratio, unlike the dietary ratio encountered in the Western diet today [32, 33]. The diet of industrialized societies is characterized by increased n-6 consumption and decreased n-3 consumption, resulting in a relative change in n-6/n-3 ratio over the last 150 years [32, 34, 35]. Some hypothesize that this discrepancy may influence the occurrence of lifestyle-related diseases. A 2002 WHO/FAO report reviewed the scientific evidence and indicated that a balanced ratio is essential to human health. However, a more recent WHO/FAO report stated no specific recommended level, given that intake of both n-6 and n-3 should be within the recommendations [11].

1.2.3 Epidemiological findings

Health effects from n-3 PUFAs were first described in 1961 after the observation of beneficial effects on cardiovascular disease [36]. Later, the low incidence of coronary heart disease in Greenland Inuits was linked to their high consumption of seafood [37]. Bang and Dyerberg summarized findings of reduced frequency of heart attacks in Norway during World War II and the tendency for long bleeding times in Greenland Inuits, and linked both observations to increased or high consumption of fish and decreased production of pro-inflammatory

prostaglandins [38]. Since those pioneering studies, the potential effects of PUFAs have been explored for a wide variety of diseases and in relation to disease prevention. A brief

presentation of general findings regarding potential effects of PUFAs on cancer, CVD and a selection of other diseases will be given here. Importantly, epidemiological findings have spurred a debate regarding PUFA effects. In a meta-analysis of randomized controlled trials and cohorts, Hooper et al. found no clear evidence of effects on general mortality, CVD events, or cancer [39]. However, the authors did not suggest changing dietary advice to increase fish or n-3 consumption. Their main critique towards findings from cohort studies pointed to the difficulty of adequately adjusting for lifestyle factors associated with fish/n-3 consumption, which generally is higher in people with a healthy lifestyle. Additional uncertainties in the available evidence is introduced by methodological issues such as short

(24)

follow-up time, the possibility of errors in estimations of intake, not considering individual PUFAs or n-6 and n-3 PUFAs separately, lack of consensus regarding adjustment for total energy intake, or biological factors like varying bioavailability of FAs and the influence of genetic polymorphisms [40-42]. The conclusions from Hooper et al. [39] were challenged in a report by a panel of experts appointed by the International Society for the Study of Fatty Acids and Lipids [43]. The authors argued that the combined weight of the scientific evidence from multiple study design types clearly demonstrate the benefit of fish oil PUFAs, especially for CVD (further discussed below).

- Cancer

Cancer causes 13% of all deaths worldwide, and is the leading cause of death in the developed world [44]. Although it is a diverse group of diseases, all cancers are characterized by

uncontrolled cell proliferation. Tumourigenesis is viewed as a multistep process, reflecting the accumulation of genetic changes that enable the transformation of normal cells into

malignant ones [45]. Importantly, tumours are no longer viewed as isolated entities of clonal cell proliferation, but rather as complex tissues composed of multiple cell types, that are actively interacting with their surroundings, known as the tumour microenvironment [46].

Via the blood stream or the lymphatic system the cancer might spread to distant parts of the body, forming metastases. Epidemiological studies exploring the influence of dietary PUFA intake on cancer incidence and/or death are numerous, however, firm conclusions have not been reached [17, 47, 48]. More consistent results have come from studies using biomarkers of exposure such as concentrations of single PUFAs in blood or plasma lipid fractions, rather than intake assessment. A protective effect was found for example for breast cancer, although the number of studies is limited [49, 50]. The WCRF/AICR report made no statements for PUFAs, because the report focused on food, not single dietary constituents. Clinical trials using defined amounts of specific PUFAs for cancer prevention or treatment (as co-treatment, or to alleviate symptoms of cachexia) are more rigorously designed, and may yield informative results [51]. Although a few trials are registered in the clinicaltrial.gov database, few results have been published [51]. More -and more carefully designed studies, are needed to explore the potential associations of single PUFAs or FA ratios on cancer incidence [42], such as the

(25)

VITAL study [52]. The epidemiological findings contrast the wealth of mechanistic evidence provided from cell culture studies that link PUFAs to several processes of carcinogenesis.

Mechanisms include modulation of cell cycle control, inflammation, oxidative stress and cell membrane structure [42], further described in chapter 1.2.4.

- CVD

CVD is a leading cause of death in the developed world. Atherosclerosis is the result of slow, long-time deposition of lipids (cholesterol and triglyceride) in blood vessels, accompanied by vessel wall thickening and inflammation, as well as accumulation of macrophages engulfing large amounts of lipids (foam cells). The atherosclerotic plaque causes reduced blood flow at its site of origin, or may rupture and cause thrombosis and infarction (cell death due to lack of oxygen) at distal sites including the heart (“heart attacks”). It has been established that the marine n-3 PUFAs EPA and DHA reduce CVD risk factors, incidents, and deaths [43, 53-55].

Although this conclusion was challenged by the meta-analysis mentioned above [39], more recent publications have supported it [11]. On the other hand, the plant-derived ALA and n-6 PUFAs have not been assigned the same effects [53, 55-57]. Mechanisms of PUFAs in relation to CVD involve the reduction of plasma cholesterol and triglyceride levels, platelet

aggregation and inflammation. The relative importance of the different mechanisms has not been established. In Norway, n-3 PUFAs are used as a prescription drug to treat

hypertriglyceridemia, and as co-treatment for secondary prevention after heart attacks [30].

- Other diseases

PUFAs have been implicated in the etiology of several other diseases, primarily those involving an immunological aspect. Autoimmune diseases like rheumatoid arthritis may be reduced due to the modulation of AA-mediated inflammatory responses. Although some studies, including randomized controlled trials, have shown positive effects on chronic inflammatory diseases, the data has not been judged as adequate to inform clinical

practice [58]. High levels of PUFAs or high fish consumption have been linked to increased risk of type 2 diabetes mellitus [59]. However, a systematic review/meta-analysis found no such associations, but concluded that the plant-derived ALA may be associated with modestly

(26)

lower risk of type 2 diabetes mellitus [60]. The WHO/FAO report concluded that there is possible evidence for a reduced risk of diabetes with increased PUFA intake [11]. Collectively, the conclusions are conflicting. Lastly, studies on PUFA exposure and risk of atopy/asthma in children and adolescents have shown promising results [61].

1.2.4 Molecular mechanisms

The specificity inferred by the chemical structure of PUFAs is a key aspect to their differential influence on physiological, cellular, and molecular mechanisms. The (relative) abundance of different FAs has been linked to disease prevention and pathogenesis as outlined above, but despite more than 30 years of research, the mechanisms of action are not fully understood.

Hypotheses linking PUFAs to disease prevention and treatment is to a large extent related to their function as immunoactive substances [62]. Inflammation is essential to maintain health and homeostasis, but pathological inflammation may occur if regulatory control is lost, and it may contribute to the pathogenesis of chronic diseases and cancer. It has been firmly

established that major inflammatory signalling pathways are modulated by PUFAs, but they do so in a variety of different ways [62, 63].

Dietary PUFAs are incorporated into the phospholipids of cellular membranes, where they serve as precursors for lipid mediators (the eicosanoids, Figures 3 and 4). Eicosanoids work in an auto- and paracrine manner to induce signal transduction via specific cell surface

receptors. In general, lipid mediators derived from n-3 PUFAs are regarded as less pro-

inflammatory or less biologically active compared to those derived from n-6 PUFAs, although this is a simplification [64]. The identification of the prostaglandins was the basis for the 1982 Nobel Prize for Physiology or Medicine, awarded to Bergström, Samuelsson and Vane [65]. In addition to eicosanoid production, PUFAs in the membrane phospholipids influence the physical properties of the membrane and lipid rafts, for example by increasing fluidity

(Figure 4). This may affect signal transduction via membrane receptors, and the production of second messengers like diacylglycerol [64]. PUFAs interact with membrane-bound or

intracellular receptors, some of which function directly as transcription factors. Transcription factors with PUFA and/or eicosanoid ligands include the peroxisome proliferator-activating

(27)

receptors (PPARs), retinoid x-receptor (RXR), Toll-like receptors (TLR), and others [58, 66].

Ligands for PPARs include dietary fats, prostaglandins, and oxidized lipids [67]. Upon ligand binding, PPARs heterodimerize with RXRs and bind to deoxyribonucleic acid (DNA) to control expression of target genes. On the other hand, TLRs reside in cellular membranes, and ligand binding causes TLRs to dimerize and recruit adaptor proteins to transduce signals intracellularly [68]. The dimerization and recruitment steps may be influenced by the properties of the membrane lipid bilayer [68]. Collectively, PUFAs may influence cellular processes like cycle regulation, apoptosis, endoplasmic reticulum stress, autophagy, lipid metabolism, oxidative stress, and calcium homeostasis [24, 69, 70].

Figure 4: Major pathways of fatty acid production, transport, and metabolism. Full names are stated in the list of abbreviations. Reprinted with permission [71].

A multitude of mechanisms that modulate the activity of immune cells may be influenced by dietary PUFAs. Many of these, if not the majority of them, will –directly or indirectly- influence the level of gene expression in a number of inflammatory pathways. Via the mechanisms outlined above, PUFAs have been associated with expression levels of pro- inflammatory markers like TNFα, IL-1β and IL-6 in cell lines [62]. These findings have to a certain degree been verified in humans. For example, fish consumption was associated with decreased levels of biomarkers of endothelial dysfunction and low-grade inflammation in healthy adults [72]. The same tendency was shown for gene expression levels in an

intervention study to reduce the n-6/n-3 ratio [73]. Two recent reviews examined the

(28)

association of inflammatory markers and n-3 PUFAs in individuals who were healthy or at risk of CVD: one concluded that PUFAs reduce inflammatory markers [74], whereas the other made no such conclusion [75]. In summary, molecular mechanisms affected by PUFAs may influence several aspects of immune cell function, with possible implications for disease prevention and progression.

1.3 Vitamin D

1.3.1 Structure and metabolism

The term “vitamin D” is used to describe a group of fat-soluble secosteroids [13, 76]. There are two main sources for vitamin D: the diet, and cutaneous synthesis after ultraviolet B (UVB) exposure (Figure 5). Dietary sources, or cutaneous conversion of previtamin D, provide the prohormone vitamin D3 (cholecalciferol). Vitamin D3 enters the circulation bound to vitamin D binding protein (DBP). Vitamin D3 may be stored in adipose tissue, or taken up by the liver where it is hydroxylated to 25(OH)D. This metabolite is released to the circulation again, before being hydroxylated to the biologically active 1,25(OH)2D (calcitriol, Figure 6) in the kidney. Conversion to the active metabolite in the kidney is regulated by parathyroid hormone and fibroblast-like growth factor-23, in response to serum calcium and phosphate levels [13]. Vitamin D3 can also be converted to the active metabolite by cells of the immune system [77]. 1,25(OH)2D binds to the vitamin D receptor (VDR), and molecular functions are presented in chapter 1.3.4. A final conversion in the kidney or intestine yields the inactive calcitoic acid, which is excreted in the bile.

(29)

Figure 5: Overview of vitamin D metabolism. Full names are stated in the list of abbreviations.

Figure 6: The structure of calcitriol (1,25(OH)2D, PubChem CID 5280453). Reprinted with permission [22].

1.3.2 Sources and recommendations

The relative amount of circulating 25(OH)D that arises from cutaneous synthesis versus dietary intake has not been specified [13]. Natural dietary sources of vitamin D include fatty fish, fish oils and egg yolk. In addition, several foods are fortified with vitamin D, especially dairy products. The Norwegian recommendation for intake of vitamin D is 7.5 μg per day [15]. The recommended daily dose of cod liver oil (5 ml) provides 10 μg vitamin D, and other supplements, like multi-vitamin products, may provide significant amounts as well [12].

When excluding users of (any) dietary supplements (58% of women), mean daily dietary

(30)

intake in Norwegian women was estimated to 5 μg, and the mean intake was doubled when including users of dietary supplements [15].

When it comes to circulating 25(OH)D concentration, the contribution of cutaneous synthesis relative to dietary intake may vary substantially. For example, in the summer, the cutaneous synthesis is a significant contributor, but in the winter months it is much less so [78]. The degree of UVB exposure is influenced by a number of factors in addition to season, including weather conditions, latitude, skin pigmentation, use of clothing/sun screen, genetic factors, and age [13]. Importantly, in Norway, the sun is low or absent during

wintertime, resulting in a prolonged period of no cutaneous vitamin D synthesis [79]. It follows from this that the dietary intake is all the more important during this period to maintain sufficient circulating 25(OH)D concentration [78]. UVB assessment requires tailored questionnaires or models in epidemiological studies [79], and these are not always used. The difficulty of UVB assessment complicates the establishment of dose-response relationships for the influence of both dietary intake and cutaneous synthesis on circulating 25(OH)D.

There is an ongoing debate regarding the target concentrations of circulating 25(OH)D, and the definitions of vitamin D status based on this concentration [80, 81]. The cut-offs directly influence the estimations of vitamin D status in a population, and subsequently may strongly influence dietary advice given to the public to prevent deficiency. At the individual level, cut- offs are used in the assessment of vitamin D status in patients, and may guide advice given by physicians to prevent or treat deficiency, often using supplements. Hence, definitions of vitamin D status based on circulating 25(OH)D will impact both public and personal health care. The IOM advocates a modest target level of circulating 25(OH)D. In their extensive report on dietary reference intakes for calcium and vitamin D, which serves as advice to the U.S. government, the IOM recommended a 25(OH)D concentration of 50 nmol/L, and defined vitamin D deficiency as concentrations below 20 nmol/L [13]. However, others use 50 nmol/L as a cut-off for vitamin D deficiency, and recommend a concentration of at least 75 nmol/L [82]. The daily requirement to reach target concentrations of 25(OH)D may vary

(31)

in different population strata. For example, the elderly may require higher intake. Also, because vitamin D is not readily mobilized from storage in adipose tissue, obese people may require higher doses compared to normal weight people [13].

1.3.3 Epidemiological findings

Low sun exposure and low vitamin D intake was causally associated with rickets more than a century ago [83]. In Northern Norway, a high prevalence of rickets was found in areas with low fish consumption as early as 1931 [84]. Since then, its association with numerous health effects has been explored. Findings remain controversial, and some advocate that the non- skeletal health effects of vitamin D may be more important than the skeletal effects [85]. The IOM report stated that available scientific evidence does not support a causal relationship between vitamin D and non-skeletal health outcomes [13]. Not many randomized controlled trials have been carried out, and the many observational studies are prone to uncertainties.

These uncertainties are introduced by methodological variability (lack of standardized measurement methods, assessment of sun exposure, etc.), lack of defined dose-response relationships, and the fact that studies often co-administer vitamin D and calcium. Lastly, genetic variants in the VDR has been shown to modify the association of vitamin D and health outcomes [86]. A meta-analysis of observational studies in general populations found an inverse association between 25(OH)D and all-cause mortality [87], which was further supported by findings in a recent population-based cohort [88]. The current state of the evidence for cancer, CVD and a selection of other diseases will be briefly presented here.

- Cancer

The WCRF/AICR report on cancer prevention judged the evidence for foods containing vitamin D as “limited-suggestive” for a decreased risk of colorectal cancer [1]. This is in line with a more recent meta-analysis of intake and circulating 25(OH)D in relation to colorectal cancer [89]. The IOM report stated that evidence for colorectal cancer from both

observational studies and randomized trials is available, although limited [13]. For other cancers there is a tendency for positive findings in smaller studies using larger doses of vitamin D supplements [90], and no associations found in larger studies using lower

(32)

doses [91]. A recent meta-analysis could find no evidence to draw conclusions regarding vitamin D supplements for cancer prevention [92].

- CVD

Observational studies have indicated a possible role for vitamin D in CVD, but these studies have often been of insufficient quality [93, 94]. The IOM report judged the evidence for vitamin D in relation to CVD as not convincing, and concluded that the role of vitamin D in CVD is at present unresolved [13]. Carefully designed randomized, controlled trials are needed to further explore the relationship between vitamin D and CVD [95].

- Other diseases

Paralleling its association with rickets in children, 25(OH)D has been used as a biomarker for osteomalacia and osteoporosis in adults. However, in a recent publication pointed out several uncertainties regarding the well-established link between bone disease and vitamin D [83].

First, low 25(OH)D has not been consistently found in people with bone mineralization defects, nor is osteomalacia found in as many osteoporosis patient as would be expected if the association was causal. Second, as pointed out previously, only a few large-scale randomized controlled trials of vitamin D supplementation without simultaneous calcium

supplementation have been carried out. Meta-analyses of such studies show inconsistent results [13, 83, 96]. On the other hand, the evidence for moderate-dose vitamin D+calcium supplements to promote bone health has been judged as adequate [13, 83, 92]. The evidence for immune function, autoimmune disorders, diabetes, metabolic syndrome, physical performance including falls, and infections were evaluated at the same level as that for CVD:

not convincing, and not adequate to conclude regarding the role of, or potential benefit from, vitamin D [13].

1.3.4 Molecular mechanisms

The lack of clear epidemiological findings contrast the amount of mechanistic data that describes the effects of vitamin D at the cellular level: based on gene expression profiling, it

(33)

has been estimated that 1,25(OH)2D may influence expression levels of up to 5% of the genome [76]. Calcium and phosphate metabolism are the classical, well characterized mechanisms influenced by vitamin D. In addition, several non-classical mechanisms have emerged. These include cell growth and differentiation, apoptosis, cell adhesion, oxidative stress, DNA repair, and possibly autophagy [76, 97].

The biologically active 1,25(OH)2D is mainly synthesized in the kidney, but may also be synthesized in many other tissues and cells, including immune cells. In target cells, activated 1,25(OH)2Dbinds to the vitamin D receptor (VDR). The VDR is a transcription factor which upon ligand binding recruits its dimerization partner RXR and other co-activators, and modulates the expression of genes that contain the VDR responsive promoter

element (VDRE) [76]. 1,25(OH)2Dmay also repress gene transcription, via recruitment of co- repressors and inhibition of negative VDREs that are transcriptionally active in the absence of the ligand [76]. In addition to gene expression regulation, 1,25(OH)2Dhas been shown to influence cellular function by non-genomic mechanisms. These mechanisms are initiated by 1,25(OH)2Dinteraction with VDR located in caveolae lipid rafts of the plasma membrane.

Downstream signaling occurs via signal transduction pathways including G-protein coupled receptors, phospholipase C (PLC), mitogen-activated protein (MAP-) kinase,

phosphatidylinositol-3-kinase (PI3K) [76, 98]. These pathways may ultimately modulate gene expression levels. The specificity of genomic versus non-genomic actions is dictated by the stereochemical conformation of 1,25(OH)2D[98].

The effect of vitamin D on cells of the immune system has been recognized for more than 20 years, but the mechanisms are still under investigation. Vitamin D may modulate the

expression of several immune-regulatory compounds, such as chemokines, interleukins and adhesion molecules [99]. Cells of the adaptive immune system are generally inhibited by 1,25(OH)2D. For example, T cell proliferation, activity, and cytotoxicity may be decreased, but suppressive T regulatory cell numbers may be increased, which is important for self-

tolerance [100]. Similarly, B cell proliferation and immunoglobulin production is decreased (which may be an indirect effect, as VDR expression in B cells has not been clearly

(34)

established) [77]. Cells of the innate immune system may also be inhibited by 1,25(OH)2D,

and the induction of self-tolerance by dendritic cells may be favored [77, 101]. On the other hand, proliferation and differentiation of monocytes and macrophages may be increased, leading to a hypothesized role for vitamin D in defense against bacterial infection [77, 101].

The immune dampening effects of vitamin D have led to the investigation of potential roles in inflammatory, hypersensitive, and autoimmune diseases. Also, because inflammation is a prominent feature of other major diseases, there are possible implications of vitamin D for both cancer and CVD. Vitamin D influences the immune system, and thereby may modify inflammation as a cancer risk factor. Furthermore, direct effects on cancer cells have been clearly demonstrated [102]. There are tissue specific effects, but overall vitamin D seems to promote differentiation and reduce cancer cell growth rates by inducing cell cycle arrest or cell death [102]. In relation to CVD, one group identified that vitamin D inhibited foam cell formation and macrophage cholesterol uptake in diabetics, via reduced PPARγ signaling and reduced endoplasmic reticulum stress [103, 104]. In summary, despite the clear effects in cell line and animal studies, the physiological roles of vitamin D remains unclear [77, 98, 100].

1.4 Gene expression analysis

1.4.1 Background

Gene expression analysis is based on the central dogma of molecular biology, which was first stated by Francis Crick in 1958, and detailed by the same scientist in a Nature paper in 1970 [105]. The dogma describes the directionality of the flow of biological information from the sequence of purine and pyrimidine bases in the DNA, via transcription into mRNA, and translation into the amino acid sequence of proteins (Figure 7). The dogma entails the concept of the genetic code, which describes how triplets of nucleotides specify one amino acid. Also essential is the complementarity of the double-helical structure of DNA, specified by

complementary base pairing. Collectively, these concepts form the basis for how we

understand all living things: genetic information is encoded in the DNA in units (genes) and selectively used (expressed/transcribed to mRNA) as templates for production of proteins,

(35)

which in turn are the working units of the cell. The human genome contains approx. 20 500 genes [106].

Figure 7: Central dogma of molecular biology. Full names are stated in the list of abbreviations.

Microarray technology is currently being used to investigate all levels of the biological information flow, from epigenetics, to DNA, RNA (protein-coding and non-

coding/regulatory), as well as proteins. In principle, DNA microarrays are solid substrates (for example glass slides) with attached DNA oligonucleotide probes that correspond to genes (Figure 8). mRNA is extracted from biological samples, then purified, amplified, and labeled, before allowed to hybridize by complementary base pairing with the oligonucleotides on the microarray. After washing away unbound material, mRNA abundance is quantitated by image analysis.

(36)

Figure 8: Basic principle of hybridization-based microarrays.

The complete collection of mRNA molecules present in a certain cell or tissue at a given time is referred to as the transcriptome, and assessment of the transcriptome is called

transcriptomics. High-throughput (‘omics) technologies, including microarrays, are increasingly being used to elucidate the complex interactions between environmental exposures and human health and disease. Gene expression microarrays are used for three main purposes: 1) identification of differentially expressed genes, 2) class discovery (grouping of samples based on gene expression profiles), and 3) class prediction (assignment of new samples to pre-defined groups). In the present thesis, microarrays were used to identify differentially expressed genes according to lifestyle and technical factors (paper I), as well as PUFA ratios (paper III) and vitamin D (paper II). Paper II and III thus fall within the field of nutrigenomics, where the use of genomic technology is merged with nutritional research.

1.4.2 Challenges

There are several challenges related to transcriptomics research: pre-analytical issues, the technology itself, data analysis, and lastly, interpretation of results. Problems reported from early microarray studies included limited reproducibility and comparability between platforms, which generated skepticism against use of the technology [107]. One of the most extensive evaluations of microarray variability stemming from intra-site, inter-site and inter-

(37)

platform differences was conducted by the MicroArray Quality Control (MAQC) project, initiated by the United Stated Food and Drug Administration. The project involved both the scientific community and commercial microarray providers, and came about to shed light on the reported problems. Results of the MAQC project were initially presented in a series of articles in Nature Biotechnology in 2006 (www.nature.com/nbt/focus/maqc). The project reported that the overall technical performance of microarrays “… supports their continued use for gene expression profiling in basic and applied research…” (quote from Shi et al. [108]).

Although some of the assessment methods and results of the MAQC project have been challenged by others [109], the main conclusion has not.

One of the major challenges to providing reproducible microarray data involves the lack of consensus regarding data analysis methods. Data analysis is carried out in multiple

consecutive steps, such as preprocessing and normalization, quality control, handling of flagged spots, data filtering (the removal of transcripts or microarrays due to quality control issues), imputation of missing values, identification of differentially expressed genes, and pathway-level analysis. Each step involves making choices regarding methods, cut-offs etc., which will undoubtedly influence the reported results [110]: selecting a different method may result in a different list of genes. It has been estimated that there are 10 million possible

combinations of methods for analyzing a given microarray data set [110].

At the basis of the interpretation of microarray data was the simplistic idea of a linear coherence throughout the biological information flow (Figure 7): a gene is being expressed, and the abundance of the mRNA molecule corresponds to the amount of the translated protein, and lastly to the level of protein activity. The idea may be further extrapolated to the pathway level, and ultimately to physiological processes. Of course, this scheme far from reflects the complexity of the biological information flow, and the regulation of physiological processes through that flow. This will be further elaborated upon in the Discussion.

(38)

1.5 Human beings as a model system

There is an increasing understanding of the limitations of using animal models and cell lines to explore human pathogenic processes. Animal models have been extensively used in basic research, to perform experiments and test hypotheses that are not feasible or ethically acceptable to test in humans. However, a recent publication found that the degree of match between studies of transcriptomic inflammatory responses in humans and mice is close to random [111]. This discrepancy has also been shown for other aspects of immunology [112], and is highly relevant for the research on inflammatory processes involved in several major pathologies. However, costs for performing in-depth molecular analyses on human samples have dropped, and eligible biobanks of human sample material are being build. As a benefit of this development, the use of human sample material is increasing in biomedical research. By monitoring exposures using questionnaires and biomarkers, populations may be divided into comparison groups and thereby be regarded as “natural experiments” for the investigation of diverse arrays of exposures. This approach was employed in the present thesis.

1.5.1 Blood as a target tissue

Studies of blood cells are particularly useful for combining epidemiological and molecular aspects of biomedical research, because blood sample collection is relatively non-invasive and feasible in epidemiological cohorts. As the body’s transport system, the blood interacts with all tissues, and blood cells are exposed to nutrients, metabolites, excreted factors and waste products. Hence, the blood has been suggested as a “surrogate tissue” for the study of diseases in other tissues [113]. Furthermore, the transcriptionally active blood cells are key

immunologic regulators, and are in them self highly relevant for monitoring health status and studying inflammatory processes that are relevant for a range of pathologies including cancer and CVD.

Challenges related to gene expression profiling of blood cells are numerous. First, the blood is a complex tissue of fluctuating numbers of cell population subtypes, and cells at varying developmental stages [114]. The white blood cells (leukocytes) constitute only 1% of the blood volume, but they are the most transcriptionally active cell subtypes. Leukocytes include the

(39)

lymphocytes (20-50%, the T cells and B cells), monocytes (1-5%), and granulocytes (65%, the neutrophils, eosinophils and basophils). Platelets are cell fragments which may contain RNA from the bone marrow megakaryocytes that they originated from. Lastly, red blood cells constitute 44% of the blood volume, but they are regarded as transcriptionally inactive. The immature red blood cells (reticulocytes) are also regarded as transcriptionally inactive, but they do contain large amounts of mRNA coding for globin. This may constitute the majority of the total mRNA isolated from blood, and may pose another challenge to blood

transcriptomics by masking less abundant transcripts. Globin reduction methods are available to increase the analytical sensitivity, but the methods may introduce bias [115]. Lastly, the stabilization of the RNA profile in a blood sample directly after blood collection is necessary, as the blood cells will react to the changed surroundings by altering their gene expression and thereby occlude the relationship between expression profiles and the biology/exposure of interest.

(40)

2 Aims

The overall aim of this thesis was to assess the feasibility of performing gene expression profiling in an epidemiological setting. In more detail, we aimed at exploring the specific influences of a range of exposures on gene expression profiles in a cross-sectional study of free-living, middle-aged Norwegian women. Based on this overall aim, three questions were addressed:

1) What factors related to inter-individual differences, major exposures, and technical variation affect gene expression profiles (paper I)?

2) How do dietary FAs affect gene expression profiles (paper III)? More specifically, what gene expression differences are related to high versus low plasma ratios of LA/ALA, AA/EPA, and total n-6/n-3?

3) How does vitamin D status affect gene expression profiles (paper II)? More specifically, what gene expression differences are related to sufficient versus deficient vitamin D status, measured as plasma 25(OH)D?

4) To what degree do the measured levels of vitamin D and PUFAs correlate (paper IV)? This needs to be clarified in order to discuss the gene expression profiles related to the two

nutrients.

(41)

3 Material and methods

3.1 The Norwegian Women and Cancer Study

The Norwegian Women and Cancer Study (NOWAC, Kvinner og Kreft) was started in 1991, to investigate risk factors for breast cancer in a prospective cohort design. The study has been described in detail elsewhere [116, 117], and further information can be found online at http://site.uit.no/nowac. Women were invited to participate based on random sampling from the Norwegian Central Person Register, and enrollment has been carried out in three main steps (1991, 1995-97, 2003-07). Subjects participating for the second time were invited in 1998-2002, and those participating for the third time were invited in 2003-05. Today, the study includes more than 172 000 women aged 30-70 years at enrollment, who have answered questionnaires regarding lifestyle, health, medication use, diet, and other factors. The

questionnaires used have varied in length over the years, ranging from two to eight pages.

Furthermore, NOWAC holds an extensive biobank of more than 60 000 blood samples, as well as both cancerous (400) and healthy (400) breast tissue biopsies. Additional important features of the study includes repeated collection of exposure information (up to three questionnaires answered by each woman), and the opportunity to link to the Cancer Registry of Norway and the to the Cause of Death Registry. The required ethical approvals have been obtained from The Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate. All participants have given written informed consent.

3.1.1 The NOWAC Post-genome Cohort

The introduction of high-throughput ‘omics technologies in the past two decades, has highlighted the need for epidemiological study designs and biobanks that could take

advantage of the technological advances. The NOWAC Post-genome Cohort and its biobank was designed not only to meet the epidemiological goals of minimal error and bias (by

optimizing the selection of study population and the study population size), but also to meet the requirements of the ‘omics technologies. These requirements include adequate procedures for sample collection, shipment, storage and handling.

Blood gene expression, lifestyle and diet The Norwegian Women and Cancer Post-genome Cohort

Blood gene expression, lifestyle and diet

The Norwegian Women and Cancer Post-genome Cohort

Karina Standahl Olsen

A dissertation for the degree of Philosophiae Doctor

Blood gene expression, lifestyle and diet

The Norwegian Women and Cancer Post-genome Cohort

Karina Standahl Olsen

Department of Community Medicine Faculty of Health Sciences

University of Tromsø Tromsø, Norway

2013

Acknowledgements

Table of Contents

Summary

Sammendrag

List of papers

List of figures and tables

Abbreviations

1 Introduction

2 Aims

3 Material and methods